Communities living alongside wildlife in southwestern Nepal have spent generations deciphering the soundscape of the forest to stay safe from large predators. Now, conservation scientists have tested an automated system that applies machine learning to these traditional warning signs, detecting prey alarm cries to map the movements of wild tigers.
The trial took place near Bardiya National Park, where the national tiger count has expanded to 429 animals, including 112 residing in Bardiya alone. As populations rebound, dangerous interactions have risen sharply. Research indicates that tigers attacked 245 people across Nepal from 2010 to 2024, resulting in 141 deaths. Around Bardiya specifically, big cats caused 18 deaths and injured 10 people between 2019 and 2023.
Automating forest warnings
Rather than attempting to monitor stealthy predators directly, field biologists focused on acoustic reactions from prey species. Tests using realistic predator models demonstrated that spotted chital deer, rhesus macaques, and gray langurs reliably produce vocal alerts when large carnivores approach. Researchers then programmed compact, solar-powered listening devices, priced at approximately sixty dollars each, to detect chital deer vocalizations.
Ten monitors were positioned throughout Dalla Community Forest over a three-month trial. Whenever a deer cried out, the system beamed notifications across a wireless network, generating real-time digital maps of zones where tigers or leopards were likely active. Local forest guards verified that these automated alerts closely mirrored tracks and physical evidence found on trails, enabling managers to advise villagers to avoid specific patches while gathering fodder or firewood.
Implementation hurdles
Despite warm reception from nearby residents and forest guards, the initiative did not transition into permanent operational use in Nepal. Investigators noted technical challenges, such as overheating units, occasional false alerts, and instances where deer fled silently without calling. Furthermore, the acoustic signatures cannot differentiate between tigers and leopards.
The initiative ultimately concluded its initial phase due to shortages in operating funds, local university partnerships, and long-term staffing support, prompting researchers to transfer the monitoring equipment to India for further testing. Forestry officials in Nepal stressed that automated tools require comprehensive evaluation before wide adoption, while conservationists maintain that listening networks offer valuable potential to avert lethal encounters before they happen.
